Article

Long-Term, Stochastic Editing of RegenerativeAnatomy via Targeting Endogenous BioelectricGradients

Fallon Durant,1 Junji Morokuma,1 Christopher Fields,2 Katherine Williams,1 Dany Spencer Adams,1 andMichael Levin1,*1Allen Discovery Center at Tufts University, and Department of Biology, Tufts University, Medford, Massachusetts; and 2IndependentResearcher, Sonoma, California

ABSTRACT We show that regenerating planarians’ normal anterior-posterior pattern can be permanently rewritten by a briefperturbation of endogenous bioelectrical networks. Temporary modulation of regenerative bioelectric dynamics in amputatedtrunk fragments of planaria stochastically results in a constant ratio of regenerates with two heads to regenerates with normalmorphology. Remarkably, this is shown to be due not to partial penetrance of treatment, but a profound yet hidden alteration tothe animals’ patterning circuitry. Subsequent amputations of the morphologically normal regenerates in water result in the sameratio of double-headed to normal morphology, revealing a cryptic phenotype that is not apparent unless the animals are cut.These animals do not differ from wild-type worms in histology, expression of key polarity genes, or neoblast distribution. Instead,the altered regenerative bodyplan is stored in seemingly normal planaria via global patterns of cellular resting potential. Thisgradient is functionally instructive, and represents a multistable, epigenetic anatomical switch: experimental reversals of bioelec-tric state reset subsequent regenerative morphology back to wild-type. Hence, bioelectric properties can stably overridegenome-default target morphology, and provide a tractable control point for investigating cryptic phenotypes and the stochas-ticity of large-scale epigenetic controls.

INTRODUCTION

Regeneration is thought to be a deterministic, robust processthat precisely rebuilds the species-appropriate anatomicalstructures that are missing after significant injury (1). Herewe show a striking counterexample illustrating that physio-logical manipulation can stably alter an animal’s targetmorphology (the anatomical configuration that, oncereached, causes regenerative growth and extensive remodel-ing to cease). Transformative advances in regenerative med-icine will require understanding of the control systems thatdetermine a given organism’s large-scale anatomicalpattern. Such controllers must integrate mechanisms overmultiple scales, from molecular-level biochemical pro-cesses to system-level anatomical outcomes (2,3). Touncover biophysical epigenetic controls of body-widepatterning, we investigated the role of endogenous bioelec-tric signaling in this process in the highly regenerativeplanarian (Dugesia japonica).

Submitted January 20, 2017, and accepted for publication April 14, 2017.

*Correspondence: michael.levin@tufts.edu

Editor: Stanislav Shvartsman.

http://dx.doi.org/10.1016/j.bpj.2017.04.011

� 2017 Biophysical Society.

This is an open access article under the CC BY-NC-ND license (http://

creativecommons.org/licenses/by-nc-nd/4.0/).

Endogenous bioelectricity is an important emerging field(4–10). Numerous studies over the last century have charac-terized endogenous electric fields and ion currents thatprecede and predict important events in embryogenesis,regeneration, and cancer (10–19). Functional work hasimplicated developmental bioelectric states in the regulationof cell behavior such as proliferation, apoptosis, and differ-entiation (20,21), as well as tissue-level responses suchas wound healing (22–25) and appendage regeneration(26–34). Recent advances have resulted in the establishmentof molecular-genetic tools for manipulation of bioelectricsignaling (35–38), and the integration of this biophysical,epigenetic layer of control with canonical signaling path-ways and downstream transcriptional cascades (16,39–41).This field is also of significant importance for biomedicine,via the identification of numerous developmental channelo-pathies (42–44), the application of bioelectric signaling incancer (36,45–48), and an appreciation of its role in stemcell bioengineering (49–53). It is becoming clear thatendogenous voltage gradients are powerful, instructivecues that regulate growth and form (54). However, keyquestions remain about the bioelectric code—the mapping

Biophysical Journal 112, 2231–2243, May 23, 2017 2231

Durant et al.

between stable (nonneural) electric states and subsequentanatomical outcomes.

Classical data in this field (55–58) addressed the ability ofexternal electrical stimulation to reset the polarity of frag-ments in planaria—a system with extremely robust regen-erative patterning. Recently, our group used biophysical(59–61) and computational (62–66) approaches to probethe fundamental issue of how pieces of planaria know whichstructures to make at which end. Here, we specifically testthe hypothesis that bodywide bioelectric gradients serve asa kind of pattern memory (67)—an input into the integratedcellular decisions that allow the animal to recreate the same,invariant target morphology after damage and stop growthand remodeling as soon as that correct anatomy has beenachieved. We also address the ubiquitous issue of variability,investigating in depth what happens to animals that seem toescape a particular experimental perturbation.

We found that altered bioelectric patterns underliediscrete, stochastic, and stable alterations to postregenera-tion axial polarity. Moreover, experimentally inverting thebioelectric signal rescued normal bodyplan anatomy,showing the bioelectric gradients to function as a bistablepattern memory switch that can reverse molecular-geneticcommitment of axial polarity. We also uncovered a crypticphenotype in animals that appear normal under molecularand histological assays but regenerate with a differentpattern when cut due to their internal bioelectric state.This reveals that the bioelectric switch stores the patternfor future regeneration events within a currently normalanatomy. These experiments reveal a powerful instructiverole for bioelectricity as a mediator of reprogramming ofregenerative morphology.

MATERIALS AND METHODS

Colony care and sample sizes

A clonal strain of D. japonica was kept and maintained as in Oviedo et al.

(68) and was starved for >7 days before cutting, a standard method used in

this model system to reduce variability of data by controlling for the meta-

bolic status of individuals (68). Planaria continued to be starved for the

duration of the full experiment, which had no detrimental effect on regen-

erative ability or speed. Planaria at the start of each experiment were be-

tween 5 and 20 mm before being amputated into fragments. For each

experiment, standard numbers of worms routinely used in the planarian

literature were utilized to sufficiently capture significant differences given

the amount of variation that is known to exist in each type of measurement.

Gap junction blockage and amputations

Amputations were performed as in Nogi and Levin (69). Fragments con-

taining the pharynx were made using a sharp scalpel on a moistened and

cooled Kimwipe (Kimberly-Clark, New Milford, CT) and piece of filter pa-

per. Within 30 min of cutting, pharynx-containing fragments were trans-

ferred into an octanol solution (8-OH), prepared by slowly diluting 10 mL

of 1-octanol (RM00050; Sigma-Aldrich, St. Louis, MO) into 500 mL of

commercial natural spring water (Poland Spring, PS; Poland Spring Water,

Framingham, MA) to a final concentration of 127 mM as in Oviedo et al.

2232 Biophysical Journal 112, 2231–2243, May 23, 2017

(70). 8-OH was left on a stir plate to mix for at least 30 min before treat-

ment. 8-OH was remade and replaced daily for the first three days

postamputation, then planaria were moved to PS in deep-dish plates

(100 � 20 mm; Fisherbrand; Thermo Fisher Scientific, Waltham, MA) to

regenerate in groups of 30 worms until day 21 before scoring and sorting.

Sample sizes represented in text reflect pooled data from multiple replicates

over the course of several months. Planaria were left at 20�C for the first

seven days, then moved to 10�C to prevent fissioning.

Phenotype scoring

Scoring was performed under a model No. SV6 dissecting microscope (Carl

Zeiss, Oberkochen, Germany). Criteria for a double-head (DH) phenotype

were at least one eye on each side of the planarian. Criteria for a cryptic

phenotype include at least one eye on the anterior side of the planarian,

no eyes on the apparent posterior region, and no indication of other ectopic

structures. The very rare (2.1%) regenerates that did not meet these scoring

requirements developed an eyeless, radial body plan and were not analyzed

further.

Secondary and tertiary amputations

Planaria were sorted according to phenotype after 21 days of regeneration

and cut on the same cutting surface using the same tools as the first round of

amputations. DH were amputated on either side of the two pharynxes.

Heads and tail tips of similar size were removed from either side of the

planarian and the remaining pharynx-containing fragments were allowed

to regenerate in deep-dish plates until day 21 in PS until the next round

of scoring, sorting, and subsequent amputation. Sample sizes represented

in text reflect pooled data frommultiple replicates over the course of several

months. These experiments focus on the pharyngeal fragment rather than

the pretail fragment found to produce 100% DHs in Oviedo et al. (70).

In situ hybridization

Animals were fixed in Carnoy solution for whole-mount in situ hybridiza-

tion as in Nogi and Levin (69) for mature anterior tissue markers. For other

probes, animals were formaldehyde-based fixed as in Pearson et al. (71).

Probes used were: CNS marker PC2 (72), tail marker Frizzled-T (73), ante-

rior markers 0821_HN, 1008_HH (74) (kind gifts from T. Gojobori), and

anterior specification marker ndk (75).

Immunofluorescence and immunohistochemistry

Animals were fixed in Carnoy solution as in Nogi and Levin (69). For

immunofluorescence, planaria were processed and imaged as in Reddien

et al. (76).

Primary antibody: a-phosphorylated histone H3 (H3P), 1:250 (Upstate

Chemical, Simpsonville, SC).

Secondary antibody: HRP-conjugated anti-Rabbit with TSA-Alexa568

anti-HRP (Molecular Probes, Eugene, OR).

Statistical analyses for immunofluorescence andimmunohistochemistry

Tail-marker analysis: Expression levels were measured using a 450-pixel-

area rectangular box drawn using FIJI software. Using the measure func-

tion, average gray values were recorded and compared. Statistical compar-

isons between head versus tail expression levels in wild-type (WT) and

cryptic planarian individuals were made using the software Microsoft Excel

(Microsoft, Redmond,WA) to calculate Student’s t-test (two-tailed distribu-

tion, paired samples, unequal variance). Statistical comparisons between

Bioelectric Control of Target Morphology

tail expression levels inWTand cryptic planaria were made using Microsoft

Excel to calculate Student’s t-test (two-tailed distribution, unpaired sam-

ples, unequal variance).

Neoblast distribution analysis: A selection extending to 12% of the total

length of the planarian on anterior and posterior ends was made using the

software FIJI (National Institutes of Health, Bethesda, MD) and images

were threshold adjusted to eliminate background fluorescence. Using the

count function, the number of neoblasts within the selection was recorded

and plotted over area. Area of the selection was calculated by converting

pixel area to mm2 using scale bars as reference. The neoblast density,

measured in number of neoblasts/mm2, was calculated in Microsoft Excel.

Statistical comparisons of neoblast density were made using Microsoft

Excel to calculate Student’s t-test (two-tailed distribution, unpaired sam-

ples, and unequal variance).

Membrane voltage-altering pharmacologicaltreatments and amputations

SCH-28080 (Sigma-Aldrich) treatments were performed as in Beane et al.

(60). DH were amputated on either side of the two pharynxes then soaked in

SCH-28080 as described. DMSO-treated controls had no effect and DH

phenotype remained persistent. Sample sizes were pooled from three tech-

nical replicates.

Membrane voltage reporter assay

Bis-[1,3-dibarbituric acid]-trimethine oxanol (DiBAC4(3); Invitrogen,

Carlsbad, CA) was used as in Adams et al. (33) and Oviedo et al. (77).

Planaria were amputated to produce pharynx-containing fragments as

above; these fragments were then soaked in PS or 8-OH solution. Planaria

regenerated for at least 21 days when they were scored and sorted according

to cryptic or DH guidelines above. Whole planaria were soaked in

DiBAC4(3) solution for>30 min before imaging and were immobilized us-

ing 2% low-melting point agarose and Planarian Immobilization Chips

(78). WT and cryptic regenerates were mounted ventralized and, whenever

possible, imaged on the same chip in tandem so that direct comparisons of

bioelectric physiology could be made. Animals were tracked individually in

multiwell, non-treated cell culture plates (12-well; Falcon/Corning, Corn-

ing, NY) then reamputated following the procedure in Secondary and

Tertiary Amputations. Functionality of DiBAC4(3) as a voltage reporter

(33,60,61,77,79–84) was reconfirmed by imaging and comparing WT

worms with animals soaked in a 1 mM valinomycin þ 150 mM potassium

gluconate (valinomycin þ KGluc) depolarization solution (see Fig. S5 for

data and analysis). Voltage profiling data are limited to the outermost few

cell layers, as the depth of dye imaging is limited to, at most, �50 mm

due to opacity and pigmentation of planaria tissues.

Image collection and processing

Membrane voltage (Vmem) images were collected using a model No. AZ100

Stereomicroscope (Nikon, Melville, NY) with a model No. DL-604M VP

camera (Andor Technology, South Windsor, CT), using an epifluorescence

optics FITC filter (GFP HC: 470/40, 495, 525/50), and pseudocolored

images were created using the NIS-Elements imaging software (Nikon).

Vmem images were flat-field corrected using the software MetaMorph (Mo-

lecular Devices; https://www.moleculardevices.com/systems/metamorph-

research-imaging). All other images were collected using a model No.

SMZ1500 microscope (Nikon) with a Retiga 2000R camera (Qimaging,

Surrey, BC, Canada) and Q-Capture imaging software (Qimaging). Adobe

Photoshop (Adobe Systems, San Jose, CA) was used to organize figures,

rotate and scale images, and improve visibility of entire image with the

exception of the Vmem images. Quantitative analysis was performed using

the software FIJI, with neoblast counting using the threshold function

and bioelectric analysis using the measure function. Data were neither

added nor subtracted; original images available upon request.

Statistical analysis for phenotypic data

State diagram: The ratio of DH to cryptic morphologies across all experi-

ments was close to 25:75%. The statistical significance of departures

from the fixed 25:75% ratio shown in the state diagram (Fig. 5) was

made using Microsoft Excel to calculate Student’s t-test (two-tailed distri-

bution, unpaired samples, unequal variance).

SCH-28080: Error bars for % phenotypes were calculated using 95%

confidence intervals.

Statistical analyses for bioelectric physiology

A box with area proportional to 10% of the length of the planarian was

drawn in the anterior and posterior regions of both WTand cryptic planaria.

Using the measure function in the software FIJI, average intensity values

within the box were recorded. Statistical comparisons between WT and

cryptic tails, as well as between WT and valinomycin þ KGluc-treated

animals, were made using Microsoft Excel to calculate Student’s t-test

(two-tailed distribution, paired samples, unequal variance).

Predictive modeling

A quantitative model of the stochastic branching between cryptic and DH

morphologies across multiple rounds of regeneration was implemented us-

ing JavaScript (https://www.javascript.com/) and the HTML5 canvas func-

tion. The model can be manipulated and its source code examined at http://

chrisfieldsresearch.com/GJ-memory-model.htm. For model equations, see

Fig. S6.

RESULTS

Editing the bioelectrical network stably altersregenerativemorphology to a stochastic outcome

Planaria exhibit extensive regenerative capacity, where eachfragment regenerates precisely what is missing to completea planarian-specific anatomy (i.e., target morphology)(85–87). Given the widespread biological use of bioelectriccircuits for memory (88,89) and the control of embryonicorgan identity by endogenous voltage gradients (10,67,90),we investigated the role of bioelectric signaling in regener-ative pattern memory. Bioelectric signals are changes in thespatio-temporal pattern of slowly changing (steady-state)endogenous anatomical gradients of resting potentialsacross many tissue types (6,10). Bioelectric circuits in neu-ral and nonneural tissues employ electrochemical synapsesknown as gap junctions (GJ) (91). In planaria, gap junctionalcommunication (GJC) via innexins and specific Vmem sig-nals are required for proper anterior-posterior (A/P) polarity(60,69,70). Because planaria are not amenable to channelmisexpression technology, and because RNAi approachesdo not allow transient modulation, we exploited a pharma-cological tool that can target electrical synapses in a tran-sient manner. We utilized 8-OH, a widely used GJ blocker(92,93), to alter the dynamics of the bioelectric network inplanarian fragments. Exposure to 8-OH reduces the ability

Biophysical Journal 112, 2231–2243, May 23, 2017 2233

Durant et al.

of cells to communicate ionically, partially fragmenting thelong-range circuit and resulting in a greater number ofsmaller adjacent isopotential domains (shown and analyzedin (59)).

Treatment of trunk fragments (planaria that have hadtheir heads and tails amputated) with 8-OH disrupts A/P po-larity producing a DH phenotype, without toxicity, muta-genic effects, or changes in neoblast maintenance (70)(Fig. 1 A). 8-OH treatment phenocopies the gene-specificknockdown of three innexins (70), which also producesthe exact same biaxial DH outcome. But unlike injectionof RNAi, which cannot be removed or turned off, ourmethod allows testing for persistent effects: although it effi-ciently blocks GJC and thus effectively increases the num-ber of distinct, smaller adjacent regions of unique Vmem

(isopotential domains) (59), 8-OH is known by GC-MSanalysis to efficiently wash out of worm tissues in 24 h(70). The most remarkable aspect of the DH phenotypeinduced by a 48 h 8-OH GJC disruption, is that it is, inthe absence of further treatments, permanent: both headscan be removed, the middle fragment can be allowed toregenerate, and the heads removed again multiple times,

2234 Biophysical Journal 112, 2231–2243, May 23, 2017

over months until the worms are too small to cut, in plainwater long after the original drug is gone. Yet, the DHmorphology recurs in perpetuity (Fig. 1 B), even after spon-taneous fission in plain water—the animals’ most commonreproductive mode (see Fig. S1 B), as well as after any addi-tional subsequent rounds of 8-OH treatment (100%, N ¼100). Thus, species-specific axial pattern can be overriddenby briefly changing the connectivity of a physiologicalnetwork. This finding left several key open questions. Canthe physiological state regulate molecular-genetic machin-ery that normally establishes a tail at the posterior and,and if so, is there a short window of competence for thiseffect? What stable property mediates the informationneeded for a fragment to decide to make one head or two?Why do some animals escape the effects of 8-OH exposureand regenerate single-headed (SH)? And finally, can DHworms ever be reset back to normal? Here, we addressedthese unknowns in this fascinating model system.

Importantly, the effects of 8-OH exposure do not convertall of the animals in an experimental cohort (CRPT; Fig. 1A)—they exhibit differential responses, even thoughthey are clonal individuals. Upon treatment with 8-OH

FIGURE 1 Current morphology does not neces-

sarily represent target morphology. Representative

phenotypes following GJ blocker or water treatment

are shown. Conditions: (A) WT D. japonica pharyn-

geal fragments treated in 8-OH 21 days postamputa-

tion, 72% cryptic (CRPT), 25% DH (N ¼ 573). (B)

Twenty-one days postamputation, DH D. japonica

created from pharyngeal fragments treated in

8-OH were recut in water. Heads were completely

removed past the auricles on each side in equal

amounts, 100% DH (N ¼ 100). Left side DH trunk

fragment regenerated the DH shown on the right.

(C) Twenty-one days postamputation, D. japonica

cryptic planaria created from pharyngeal fragments

treated in 8-OH were recut in water. Head was

completely removed past the auricles on anterior

side, and an equal amount was removed from the

posterior side, 77% cryptic (N ¼ 155), 23% DH

(N ¼ 155). Left side cryptic fragment regenerated

the DH shown on the right. Right side cryptic regen-

erated from another parent cryptic planarian not

shown. (D) WT D. japonica pharyngeal fragments

after an immediate, subsequent transverse amputa-

tion can regenerate independent patterning fates

(both DH and cryptic). Scale bars represent 1 mm.

To see this figure in color, go online.

FIGURE 2 Transient GJ inhibition overrides expression of early poste-

rior markers. (A) Planaria were allowed to regenerate for 2 days in water

before they were exposed to a late 3-day treatment in 8-OH. (B) Two

days postamputation, WT D. japonica express early tail markers (FzT) at

the posterior blastema (100%, N ¼ 10). (C) Five days postamputation,

WT D. japonica continue to express tail markers at the posterior end

over a wider area (100%, N ¼ 10). (D) If planaria are treated in 8-OH

from days 2–5, after the posterior genetic program is initiated (see B),

tail marker expression is no longer visible at 5 days (30%,N¼ 10) or dimin-

ished (30%, N ¼ 10). (E) Timeline of late 8-OH treatment. Planaria are

amputated and soaked in water for two days, at which point tail markers

become visible. GJC is inhibited after tail specification using a 3-day

soak in 8-OH from days 2–5. By day 5, the decision is reversed and tail

markers are no longer visible. Scale bars represent 1 mm. n R 10 for

each. To see this figure in color, go online.

Bioelectric Control of Target Morphology

(N ¼ 593), trunk fragments produce 25% DH planaria,72% planaria with normal SH A/P polarity (CRPT), and�3% planaria with indiscriminate shapes lacking eyescompletely. At first we considered the morphologicallynormal planaria resulting from 8-OH treatment to be es-capees—those that failed to be affected by the treatment,as this kind of partial penetrance is observed with practicallyevery functional protocol in most published studies. We thenasked: were these animals really escapees, or might controlcircuits have been altered in a way that is not yet apparent?Remarkably, when morphologically normal regenerateswere amputated again, in plain spring water, as long as eightweeks after the original amputation in 8-OH, it was discov-ered that they were in fact not normal with respect to theirregenerative outcomes: 23% (N ¼ 155) regenerated as DHplanaria (Fig. 1 C) with true, duplicated anterior structures(Fig. S2). DH worms were similarly obtained from frag-ments taken posterior to the pharynx, ruling out the neces-sity of a specific body region (or one having multiplepharynxes) for producing the DH phenotype in subsequentgenerations. Thus, morphologically normal cryptic planariaclearly did not escape the effects of 8-OH, but their alter-ation is not apparent in the intact state: their altered patternmemory is revealed only upon regeneration, despite theirnormal anatomical phenotype. Thus, the SH cryptic fractionresulting from 8-OH treatment are not WT (which wouldregenerate as 100% SH) but are altered so as to generatethe same proportion of SH and DH forms upon multiple sub-sequent rounds of cutting in plain water (Fig. S3).

It has been reported that very thin fragments sometimesregenerate DH planarians in water spontaneously (94); how-ever, these fragments were of standard experimental sizenormally seen in the field, which never spontaneouslyregenerate DHs. Although biaxial DH phenotypes havebeen reported classically as so-called ‘‘heteromorphoses’’(95), and the persistence of DH memory has been shown(70), this is to our knowledge the first situation in whichcompletely normal-seeming worms will regenerate to aradically different bodyplan upon amputation in plain water:target morphology (the form to make upon future injury)can be edited to be different than current morphology,despite the normal high degree of robustness in this highlyregenerative species.

Furthermore, the phenotype is stochastic, in the sense thatdiscrete outcomes occur in predictable ratios within atreated population, but individual animals (though clonaland living in the same environment) regenerate towarddistinct outcomes despite identical experimental perturba-tion. To test whether the stochastic ratio of anatomical out-comes could be changed by a second exposure to 8-OH, aseparate cohort of cryptic planaria was generated accordingto protocol and was amputated, but this time treated in 8-OHa second time. Again, the same stochastic ratios were seen,24.5% (N ¼ 439) DH and 75% (N ¼ 439) had normalmorphology. Approximately 0.5% regenerated indiscrimi-

nate shapes without eyes (N ¼ 439). We further confirmedthis anatomical coin-toss by testing adjacent pieces of acryptic worm that came from the same A/P level. We foundthat the stochastic decision to become DH or cryptic is madeindependently by each fragment, rather than determined bysome property of the parent worm, as exhibited by the abil-ity of two transverse pieces of a WT worm fragment todevelop two completely different patterning fates after8-OH treatment (Fig. 1 D).

To our knowledge, no data have been available on the sto-chastic nature of anatomical reprogramming. One keyrequirement of such a system is that it directs downstreamactivity of the genetic programs that establish head andtail structures. We found that even after posterior identityat the tail-facing wound is established at the transcriptionallevel, anatomical specification can be reversed by subse-quent instructive physiological signals, which respecifyboth the molecular and the anatomical state of the new tis-sue (Fig. 2). More importantly, however, the new stochastic

Biophysical Journal 112, 2231–2243, May 23, 2017 2235

Durant et al.

DH pattern has to be somehow stored within the bodyof cryptic planaria, so that it can guide the results of subse-quent regeneration. What stable histological, cellular,molecular, or physiological state distinguishes true WTplanaria from normal-looking (SH) cryptic planaria?

Target morphology is not stored via maturetissues, key early anterior mRNA expression, orneoblast distribution

Cryptic worms were indistinguishable from normal wormswith respect to a number of key molecular properties, inaddition to their normal anatomy. Cryptic planaria did nothave internal mature anterior structures that could have trig-gered regeneration of a second head upon amputation. AllDH planaria (100%, N ¼ 5) exhibited strong expression inboth heads of head tissue markers including PC2 (72),0821_HN, and 1008_HH (74), which are expressed by thecentral nervous system, the sensory cells of the anteriorfringe, and brain branches, respectively (Fig. 3, C, F,and I). In contrast, cryptic planaria (100%, N ¼ 5), likeWT, only exhibited expression of anterior tissue markers

t-test) and is not statistically different from expression seen in WT (p ¼ 0.6313

the two heads was not statistically significant (p ¼ 0.4079) All expression leve

(450 pixel area, measured function of shaded values; FIJI software used). (P and

tochemistry, all planes and neoblasts visible) in the region anterior to the eyes. N

tail. (R) DH have this region devoid of neoblasts on both sides of the worm. Over

to DH despite the fact that 25% will go on to regenerate DH. Scale bars repres

2236 Biophysical Journal 112, 2231–2243, May 23, 2017

at the single anterior end (Fig. 3, B, E, and H), showingthat cryptic planaria have normal histological A/P polarity(Fig. 3, A, D, and G).

Likewise, cryptic planaria did not have aberrant expres-sion of early anterior driver genes in their posterior ends.Focusing on the most essential candidate, we examinedndk, one of the earliest anterior specification markers in theplanarian (75), commonly used in various planarian species(96–99) due to its function of regionalizing where brain tis-sues may form (75). WT planaria cut and regenerated in wa-ter express ndk robustly and only in the anterior region wherethe cephalic ganglion resides (100%, N¼ 13; Fig. 3 J). Like-wise, SH cryptic planaria also only expressed ndk in the mostanterior region (100%,N¼ 10; Fig. 3K). Positive control DHworms expressed ndk robustly at both ends (100%, N ¼ 10;Fig. 3 L). Thus, the regeneration of cryptic planaria as DHis not due to ectopic presence of early essential head-specificmRNA such as ndk, although it cannot be ruled out that dif-ferences could be found by future studies using as-yet undis-covered anterior determinant genes.

To test whether cryptic planarians’ posterior ends mightbe destabilized due to failure to express required tail

FIGURE 3 Target morphology is not stored via

mature tissues, candidate early anterior mRNA

expression, or neoblast distribution. Tissue-spe-

cific analyses were performed at 21 days postam-

putation. n R 5 for each. WT D. japonica, cut

and treated in 8-OH (127 mM), went on to regen-

erate a population of DH and a population of

cryptic planaria. These were compared to WT

D. japonica cut and treated in water as a control.

Regions showing expression pattern are indicated

using green arrows. Regions devoid of expression

pattern are indicated using pink arrows. (A and

B) WT and cryptic planaria express PC2 CNS

marker robustly in the brain and ventral nerve

cords that extend to the tip of the tail whereas

(C) DH planaria express PC2 robustly in a bipolar

fashion at each end of the worm. (D), (E), (G), and

(H) WT and cryptic planaria also express mature

anterior fringe tissues (0821_HN) and brain

branches (1008_HH) only in the head, not in the

tail, whereas (F) and (I) DH express these robustly

at each end of the worm. (J and K) WT and cryptic

planaria are also only limited to robust ndk expres-

sion throughout the head region. (L) Conversely,

DH planaria express the early anterior mRNA

ndk robustly at both ends. (M) WT planaria express

FzT tail marker significantly near the tail tip as

compared with the anterior tip (p ¼ 0.0001, two-

tailed, paired Student’s t-test). (N) Similarly,

cryptic planaria also significantly express FzT tail

marker near the tail tip as compared with the ante-

rior tip (p ¼ 0.0004, two-tailed, paired Student’s

, two-tailed, unpaired Student’s t-test). (O) DH FzT expression comparing

ls for FzT were quantitatively measured for comparison using raw images

Q) WT and cryptic D. japonica are devoid of neoblasts (H3P immunohis-

eoblasts begin immediately posterior to the eyes and extend to the tip of the

all, cryptic planaria, in terms of these markers, are more similar to WT than

ent 1 mm. To see this figure in color, go online.

FIGURE 4 Cryptic planaria have differential bioelectric physiology.

(A and B) Vmem reporter assay using DiBAC4(3). Images are pseudocolored

blue-green. Brighter pixels (green) are more relatively depolarized or posi-

tively charged on the inside of cells relative to the outside. Pixels of lower

intensity (blue) are more relatively hyperpolarized or negatively charged on

the inside of cells relative to the outside. (A) WT D. japonica; anterior cells,

particularly around the region of the auricles (orange arrow), are more rela-

tively depolarized than posterior cells. (B) Cryptic planaria also have rela-

tively depolarized anterior cells, especially around the auricles (left orange

arrow) but also have a region of relative depolarization at the tail (right

Bioelectric Control of Target Morphology

markers, we probed the expression of frizzled-T (FzT), a keyWnt receptor gene involved in specification of the posteriorregion (100); in addition to obvious qualitative differences,we also quantified the spatial distribution of the in situ hy-bridization signal. WT planaria showed the expected signif-icant difference in expression of FzT between heads andtails (p ¼ 0.0001; Fig. 3 M). Similarly, cryptic planariahad significantly higher expression of FzT in tails comparedto heads (p ¼ 0.0004; Fig. 3 N). Consistently with this, thedifference in FzT expression between the two heads of DHworms was not statistically significant (p ¼ 0.4079;Fig. 3 O). There were no statistically significant differencesin expression of FzT between WT and cryptic planaria (p ¼0.74, Fig. S2). Thus, cryptic worms’ posterior ends are truetails indistinguishable from WT tails by expression of FzT;their altered regenerative pattern is not due to a lack of tailidentity.

Neoblasts in WT and DH D. japonica are absent from theregion anterior to the photoreceptors (Fig. 3, P and R) (101).If abnormal target morphology was stored in cryptic planariavia headlike distribution of neoblasts in the posterior struc-tures, this would be revealed by a lack of proliferating cellsnear the tip of the posterior region of cryptic planaria. Thedistribution of mitotically active cells in cryptic tails usingthe standard H3P marker (76) did not resemble the anteriorneoblast distribution (Fig. 3 Q). In cryptic planaria, the ante-rior neoblast density significantly differs from that in theposterior tip (p<< 0.0001 for each; Fig. S4), whereas differ-ences in average neoblast density between WT and cryptictails were not statistically significant (p¼ 0.3369). This con-firms that the neoblast distribution in the cryptic worms’ tailsis not headlike. Thus, cryptic planaria do not maintain theirunique regenerative response via a head-specific distributionof neoblasts in posterior tissues.

orange arrow) that is not present in the tail region of WT. (A0 and B0) Brightfield images of animals shown in (A) and (B), respectively. Planaria were cut

according to the diagrams shown. (C and D) Resultant regenerates from

amputated planaria represented in (C) and (D). (C) A WT regenerate

produced a planarian with normal polarity after DiBAC4(3) analysis. (D)

Cryptic planaria regenerate a DH phenotype after DiBAC4(3) analysis.

(E) The DH phenotype can be reset with bioelectric manipulation. Planaria

that are DH after 8-OH treatment exhibit a permanent change in target

morphology. When soaked in DMSO upon recutting, planaria consistently

remain DH even after three generations of recuts (100%, N ¼ 102). (F) DH

planaria, recut and treated in hyperpolarizing H,K-ATPase inhibitor SCH-

28080 show a permanent reset of the target morphology back to the SH

phenotype (34%, N ¼ 102). Scale bars represent 1 mm. To see this figure

in color, go online.

Cryptic planaria have different bioelectricphysiology from normal planaria

Bioelectric prepatterns regulate cell behavior and develop-mental pattern formation (42,60,61,84,102). We hypothe-sized that cryptic planarians’ DH regenerative targetmorphology could be stored via a change in the bodywidedistribution of cellular resting potentials, especially because8-OH has been shown to alter bioelectric prepatterns bygenerating more isopotential regions of Vmem patterning inplanaria (59). Using DiBAC4(3) voltage reporter, a well-es-tablished dye used for characterization of Vmem signatures ofpatterning systems in vivo (33,59,60,77), we observed thatmature WTworms have a relatively depolarized anterior re-gion (Fig. 4 A). Crucially, cryptic planaria exhibit a second(ectopic) region of depolarization near the posterior end,including the tail (Fig. 4 B). WT and cryptic planaria wereimaged in DiBAC4(3) together in the same field to enabledirect comparison of relative Vmem signatures in a pairwiseanalysis.

After Vmem analysis, WTand cryptic planaria were bright-field imaged and recut (Fig. 4, A0 and B0). DiBAC4(3) expo-sure had no effect on WT planarian polarity (Fig. 4 C) andcorrectly indicates depolarization as verified using a depola-rizing treatment consisting of the potassium ionophore vali-nomycin and potassium gluconate (Fig. S5; (42,60,77,79–81,103,104)). Cryptic planaria that regenerated DH upon

Biophysical Journal 112, 2231–2243, May 23, 2017 2237

Durant et al.

recutting (Fig. 4D) all had shown regions of relative depolar-ization at the posterior end compared to their paired controls(p ¼ 0.025; Table S1). Similarly, unlike WT tails that aresignificantly more hyperpolarized than WT heads (p <0.01), cryptic tails are not significantly different from crypticheads (p > 0.05). These measurements specifically focusedon animals that went on to form DHs, to understand whatwas different about the worms that would change theiranatomical layout upon cutting. This significant relationshipreveals a physiological signature that, unlike the other testedproperties, reveals what changed in the cryptic planaria: theyhave a headlike bioelectric state, in at least the surface tissuesof their posterior end, which could be responsible for the DHregeneration outcome. These data show that bioelectric stateis a unique marker of altered target morphology in anatomi-cally and molecularly normal planaria, which would reveal,to our knowledge, their novel regenerative pattern uponfuture rounds of amputation.

The above data suggest a specific manipulation foraltering the encoded target morphology. To test the instruc-tive, functional role for the voltage gradient we observed,we asked: can the persistent DH morphology be rewrittenback to a normal WT condition by manipulation of Vmem

gradients? The alteration of subsequent rounds’ animals’target morphology induced by changes of bioelectric con-nectivity predicts that directed manipulation of restingpotentials (4,10,59–61,63,67,105–107) could be used toreverse the worms’ altered anatomical outcomes. Is the cir-cuit we observed a true bistable pattern memory switch?

No gain-of-function (misexpression) technology exists forplanaria; thus dominant channel misexpression (used in othermodel systems for control of developmental bioelectricity(35,108)) was not feasible. We exploited the rich pharmaco-logical toolkit that exists for modulating bioelectric state byregulating native ion conductances (34). The H,K-ATPaseion pump inhibitor SCH-28080, validated by gene-specificknockdown of H,K-ATPase by RNAi, have been shown torobustly hyperpolarize planarian tissues, as confirmed withDiBAC4(3) imaging (61); inhibit head identity in amputatedfragments; and suppress ectopic head formation even in thepresence of otherwise DH-producing b-catenin RNAi (60).However, the effects of hyperpolarization have never beenexplored in the context of permanent (multigeneration) DHphenotypes. Thus, we sought to determine whether we couldreset target morphology back to normal from the mostextreme example of revised target morphology, the persistentDH state. We observed that 34% of DH planaria (N ¼ 102;Fig. 4 F) cut and treated with SCH-28080 exhibited normalSH regeneration, which was persistent across multiple gener-ations, unlike controls in vehicle (DMSO) that continued toregenerate as DH (100%, N ¼ 102; Fig. 4 E). Similarly, afterhyperpolarization with SCH-28080, the rate of cryptic wormconversion to the DH phenotype postamputation wasreduced to 1% (N ¼ 100). In all cases, the stable outcomewas WT (SH worms that only give rise to SH worms in sub-

2238 Biophysical Journal 112, 2231–2243, May 23, 2017

sequent generations), not cryptic. These functional resultsconfirm the voltage relevance of our dye pattern: the bioelec-tric method and data correctly predicted and identified afunctional treatment that efficiently resets the DH stateback to normal.

Most crucially, this was not a transient suppression, butwas stable reprogramming: the resulting SH phenotype re-mained persistent after at least four consecutive generationsof recuts (34%, N¼ 102; Fig. S1 C). All control planaria cutand treated in vehicle (DMSO) remained DH across all gen-erations of recuts (100%, N ¼ 102; Fig. S1 A). Thus, thealtered regenerative anatomical pattern encoded in theplanarian body can be forced back to a WT state by resettingthe bioelectrical circuit—the stable voltage prepattern isfunctionally determinative of regeneration outcome.

DISCUSSION

‘‘The so-called ‘polarity’ exhibited in the regeneration of an-

imals has suggested the idea . that the phenomenon might

be related to the outcome of differences in potential in

different regions; or, in other words, of electrical polarity.

. If this relation should be found to exist, there is a further

opportunity of testing the validity of the conclusion in the

case of axial heteromorphosis.’’

—T. H. Morgan (95)

Even in genetically identical populations, individuals canexhibit heterogeneous phenotypes that could increase fitnessin changing environments (109), drive decisions regardingcell fate and lineage (110), or lead to differences in overallshape (59). Several studies (reviewed in (111)) have shownthat target morphology, or pattern memory, can be manipu-lated, but no endogenous control mechanisms are known.Here we show that transient inhibition of GJC permanentlyalters the shape to which planaria will regenerate, via a sto-chastic mechanism involving a stable bioelectric prepattern.This bioelectric prepattern exists at least in the surface tis-sues of the worm; future development of transgenesis andtissue-specific promoters in planaria will exploit geneticallyencoded voltage reporters to probe which cell types store thebioelectric state.

The outcomes of planarian regeneration in each of theobserved stable configurations that result from GJ networkperturbation are summarized in a state transition diagram(Fig. 5). The DH state is a terminal node; worms thatregenerate as DH never spontaneously go back to a crypticor WT SH state. However, DH worms can be forced backto a SH state, and the transition probability of crypticworms to DH is almost eliminated by exogenous control ofresting potential, illustrating that the observed bioelectricdepolarization inherent in cryptic worms is a functionaldeterminant of anatomical outcome. It is possible thatadditional modalities for rewriting the target morphology

FIGURE 5 Change in target morphology is stable and stochastically var-

iable. The percentage of phenotype, CRPT (cryptic/SH) or DH (double

head), was recorded from pharyngeal fragments that were treated in GJ

blocker (8-OH), H,K-ATPase inhibitor (SCH-28080), or water. The DH

phenotype remains permanent, but cryptic planarian regeneration is sto-

chastic, resulting in a 75:25 ratio of SHs to DHs in both water and 8-OH

treatments. To see this figure in color, go online.

Bioelectric Control of Target Morphology

will be uncovered in future work; it is likewise possiblethat <100% penetrant phenotypes reported throughoutthe literature might also reflect other interesting crypticeffects, which remain to be tested by challenging theescapees in subsequent experiments. The identification ofreagents that permanently rewrite the target morphology,and insight into the mechanisms underlying such a change,have until now been hampered by the stochastic, crypticnature of this change (apparent escapees) and the robustfidelity of the regenerative program in the planarianunder most manipulations. To our knowledge, these dataare the first demonstration in a metazoan organism ofsignificant changes to the body plan that are stable, sto-chastic, and induced by a brief and purely physiologicalperturbation.

These findings significantly advance our prior work.Beane et al. (60) showed that depolarization is able toinduce head formation in posterior facing wounds in oneround of regeneration; it was unknown whether this was sta-ble (permanent) or transient. In contrast, the above datashow voltage change to be sufficient to permanently reseta stable change in the target morphology (the cryptic pheno-type) rather than merely affecting the WT state. Oviedoet al. (70) showed that DH state can be permanent; however,that study (and those from all other labs) count normal-look-ing worms as merely indicative of a less-than-fully pene-trant phenotype. In contrast, we show that this insteadreveals a subtle and profound change to the animals’ patternmemory. Lastly, we now show that bioelectric change canoverride molecular tail identity even after tail marker is ex-pressed (Fig. 2).

Given that the voltage change induces complete secondheads and resets the DH phenotype to produce tails, it isvery likely that these instructivevoltage changes lie upstreamof the endogenous gene networks that are needed to build ahead (86). Bioelectric signaling operates in concert withdownstream target genes and chromatin modifications.Thus, it is likely that additional components of the DH effectmay include other epigenetic mechanisms that integrate withbioelectrics to define outcomes above the cell level. Currentmolecular models have implicated Wnt signaling as a keyregulator of A/P polarity (reviewed in (112)); however, thesemodels are currently unable to explain the multigenerationalpersistence of the DH phenotype, the nature of stochasticityor escapees of a given treatment, and the ability of targetmorphology to be reset by bioelectric manipulations.Regardless of the putative genes or epigenetic mechanismsthat may be involved, we found a way to simply bypass themolecular genetics that influence polarity by altering endog-enous distributions of resting potential. This is attractive intranslating this work to regenerative medicine becausebioelectric physiology is a much easier target for treatmentthan genetic manipulation.

We suggest a model that correctly predicts the ratio ofcryptic to DH worms for the manipulations performed(Fig. S6). This model attributes the emergence of the crypticstate to the breakdown of an organism-wide, GJ- and polar-ization-dependent axial pattern memory that continuouslyamplifies the probability of normal, WT morphology asthe dominant target morphology in the event of injury. It as-sumes that this pattern memory is implemented by cellularresponse to a signal transmitted from anterior to posterior.The consistently correct regeneration of WT animals wouldbe the result of active, ongoing suppression by this memorysignal of the stochasticity that would be observed if targetmorphologies were allowed to compete on a level playingfield. The model postulates two effects of briefly disruptingGJ function with 8-OH treatment after an amputation: 1)memory signal reception is disrupted immediately at theanterior wound by the amputation, and slowly at the poste-rior wound by the disruption of GJ-dependent signal trans-mission, leading to stochastic competition between headand tail pathways at the posterior wound; and 2) both regen-erated heads and regenerated tails are depolarized relative tothe middle of the animal. Stochastic competition betweenpathways produces DH worms, whereas the depolarizedtail prevents the WT pattern memory from being reestab-lished after regeneration, producing the cryptic phenotype.The depolarized tail of cryptic worms disrupts anterior toposterior transmission of the memory signal; hence headand tail pathways compete stochastically at both wounds af-ter further amputations. Externally polarizing the tail of acryptic worm reinitiates anterior to posterior transmissionof the memory signal, restoring WT response to further am-putations as observed. As recognized from Turing (113)onward, the implementation of memory signals such as

Biophysical Journal 112, 2231–2243, May 23, 2017 2239

Durant et al.

postulated here requires dynamic interaction, either in theform of competition between short- and long-range signals(114), resonance as implemented in many neural networkmodels (115), enzymatic amplification of reaction rates, orsome other energy-consuming process. Our current effortsare aimed at integrating this model and its stochastic dy-namics with the recently developed bioelectric simulatorenvironment (116). We are constructing a comprehensive,realistic model comprising the known physiology, gene-reg-ulatory networks, and biochemical gradients, which can beinterrogated to extract predictions about specific interven-tions that will achieve desired outcomes in vivo.

Emergent behavior of physiological networks can supple-ment the information available from a fixed genomicsequence. The brain exploited this ancient trick long ago,as neurons and astrocytes use bioelectrical networks tomaintain and rewrite stable memories not specified by thegenome but rather modified by interactions with the envi-ronment (117). We suggest a similar scheme operates in so-matic tissues (117). Future work using a combination ofcomputational modeling platforms (116) and transgenic an-imals that will allow the use of optogenetics (37,38) willfacilitate high-resolution studies of planarian bioelectric cir-cuitry, including the specific relationship of voltage statesand electrical synapses that determine network topology.The discovery of rewritable pattern memory and accessiblestochasticity of large-scale anatomical states enable futureefforts to specify the bioelectric code in detail, testing quan-titative models of anatomical decision-making that synthe-size cell-driven emergence and top-down pattern control(118,119). It remains to be seen whether the paradigms ofmemory encoding in computational neuroscience (117),and/or other conceptual approaches for understanding cellcoordination toward species-specific shapes, will be thekey that cracks this fascinating problem.

Our results suggest that it is crucial to pursue the under-standing of stable physiological circuits with developmentalendpoints, which may provide important sources of epige-netic plasticity and interact with biochemical properties en-coded by the genome. Promising avenues for regenerativemedicine may lie in manipulating bioelectric physiologyfor the reestablishment of proper morphology. Many iontransporters are already approved for human clinical use.If bioelectric physiology plays a role in storing morphology,these electroceutical drugs could have promise in reintro-ducing shape and structure that has been lost. Future workintegrating stem cell controls, biochemical networks, andphysiological circuits will enable a new depth of under-standing of the origin of anatomical form and applicationsto its rational control.

SUPPORTING MATERIAL

SupportingMaterials andMethods, six figures, and one table are available at

http://www.biophysj.org/biophysj/supplemental/S0006-3495(17)30427-7.

2240 Biophysical Journal 112, 2231–2243, May 23, 2017

AUTHOR CONTRIBUTIONS

M.L. and F.D. designed the experiments and interpreted data. F.D., J.M.,

and K.W. performed the planarian experiments. D.S.A. contributed to the

voltage imaging methodology. C.F. created and analyzed the model. F.D.,

J.M., C.F., D.S.A., and M.L. wrote the manuscript together.

ACKNOWLEDGMENTS

We thank the members of the Levin lab, especially Daniel Lobo for his aid

with Fig. 5, and many researchers in the planarian regeneration community

for useful discussions. We thank the diligent undergraduates, including Ai

Yamasaki, who have helped us tend to our worm colony, as well as Maya

Emmons-Bell and Joshua LaPalme for thoughtful discussions and support.

We gratefully acknowledge support by an Allen Discovery Center award

from the Paul G. Allen Frontiers Group (No. 12171), the G. Harold and

Leila Y. Mathers Charitable Foundation (No. TFU141), the Templeton

World Charity Foundation (No. TWCF0089/AB55), and the National Sci-

ence Foundation (NSF) (Nos. EF-1124651 and IGERT DGE-1144591).

REFERENCES

1. Birnbaum, K. D., and A. Sanchez Alvarado. 2008. Slicing acrosskingdoms: regeneration in plants and animals. Cell. 132:697–710.

2. Losick, R., and C. Desplan. 2008. Stochasticity and cell fate. Science.320:65–68.

3. Somel, M., P. Khaitovich, ., M. Lachmann. 2006. Gene expressionbecomes heterogeneous with age. Curr. Biol. 16:R359–R360.

4. McCaig, C. D., A. M. Rajnicek, ., M. Zhao. 2005. Controlling cellbehavior electrically: current views and future potential. Physiol. Rev.85:943–978.

5. McCaig, C. D., A. M. Rajnicek, ., M. Zhao. 2002. Has electricalgrowth cone guidance found its potential? Trends Neurosci.25:354–359.

6. Funk, R. 2013. Ion gradients in tissue and organ biology (Review).Biol. Sys. Open Access. http://dx.doi.org/10.4172/2329-6577.1000105.

7. Bates, E. 2015. Ion channels in development and cancer. Annu. Rev.Cell Dev. Biol. 31:231–247.

8. Pullar, C. E. 2011. The Physiology of Bioelectricity in Development,Tissue Regeneration, and Cancer. CRC Press, Boca Raton, FL.

9. Mustard, J., and M. Levin. 2014. Bioelectrical mechanisms for pro-gramming growth and form: taming physiological networks for softbody robotics. Soft Robotics. 1:169–191.

10. Levin, M. 2014. Molecular bioelectricity: how endogenous voltagepotentials control cell behavior and instruct pattern regulation in vivo.Mol. Biol. Cell. 25:3835–3850.

11. Jaffe, L. 1982. Developmental currents, voltages, and gradients. InDevelopmental Order: Its Origin and Regulation. S. Subtelny, editor.Alan R. Liss, New York, pp. 183–215.

12. Jaffe, L. F. 1981. The role of ionic currents in establishing develop-mental pattern. Philos. Trans. R. Soc. Lond. B Biol. Sci. 295:553–566.

13. Robinson, K. R. 1985. The responses of cells to electrical fields: a re-view. J. Cell Biol. 101:2023–2027.

14. Jaffe, L. F., and R. Nuccitelli. 1977. Electrical controls of develop-ment. Annu. Rev. Biophys. Bioeng. 6:445–476.

15. Levin, M., and C. G. Stevenson. 2012. Regulation of cell behavior andtissue patterning by bioelectrical signals: challenges and opportunitiesfor biomedical engineering. Annu. Rev. Biomed. Eng. 14:295–323.

16. Tseng, A., andM. Levin. 2013. Cracking the bioelectric code: probingendogenous ionic controls of pattern formation.Commun. Integr. Biol.6:e22595.

Bioelectric Control of Target Morphology

17. Lund, E. 1947. Bioelectric Fields and Growth. University of TexasPress, Austin, TX.

18. Burr, H. S., and F. S. C. Northrop. 1935. The electro-dynamic theoryof life. Q. Rev. Biol. 10:322–333.

19. Becker, R. O. 1967. The electrical control of growth processes. Med.Times. 95:657–669.

20. Blackiston, D. J., K. A. McLaughlin, and M. Levin. 2009. Bioelectriccontrols of cell proliferation: ion channels, membrane voltage and thecell cycle. Cell Cycle. 8:3527–3536.

21. Sundelacruz, S., M. Levin, and D. L. Kaplan. 2009. Role of membranepotential in the regulation of cell proliferation and differentiation.Stem Cell Rev. 5:231–246.

22. Nuccitelli, R. 2003. A role for endogenous electric fields in woundhealing. Curr. Top. Dev. Biol. 58:1–26.

23. Reid, B., and M. Zhao. 2014. The electrical response to injury:molecular mechanisms and wound healing. Adv. Wound Care (NewRochelle). 3:184–201.

24. Zhao, M. 2009. Electrical fields in wound healing-an overriding signalthat directs cell migration. Semin. Cell Dev. Biol. 20:674–682.

25. Martin-Granados, C., and C. D. McCaig. 2014. Harnessing the elec-tric spark of life to cure skin wounds. Adv. Wound Care (New Ro-chelle). 3:127–138.

26. Smith, S. D. 1981. The role of electrode position in the electrical in-duction of limb regeneration in subadult rats. Bioelectrochem. Bio-energ. 8:661–670.

27. Smith, S. D. 1974. Effects of electrode placement on stimulation ofadult frog limb regeneration. Ann. N. Y. Acad. Sci. 238:500–507.

28. Leppik, L. P., D. Froemel,., J. H. Barker. 2015. Effects of electricalstimulation on rat limb regeneration, a new look at an old model. Sci.Rep. 5:18353.

29. Borgens, R. B., J. W. Vanable, Jr., and L. F. Jaffe. 1977. Bioelectricityand regeneration. I. Initiation of frog limb regeneration by minute cur-rents. J. Exp. Zool. 200:403–416.

30. Becker, R. O. 1972. Stimulation of partial limb regeneration in rats.Nature. 235:109–111.

31. Borgens, R. B. 1984. Are limb development and limb regenerationboth initiated by an integumentary wounding? A hypothesis.Differen-tiation. 28:87–93.

32. Jenkins, L. S., B. S. Duerstock, and R. B. Borgens. 1996. Reduction ofthe current of injury leaving the amputation inhibits limb regenerationin the red spotted newt. Dev. Biol. 178:251–262.

33. Adams, D. S., A. Masi, and M. Levin. 2007. Hþ pump-dependentchanges in membrane voltage are an early mechanism necessaryand sufficient to induce Xenopus tail regeneration. Development.134:1323–1335.

34. Tseng, A. S., W. S. Beane,., M. Levin. 2010. Induction of vertebrateregeneration by a transient sodium current. J. Neurosci. 30:13192–13200.

35. Adams, D. S. 2008. A new tool for tissue engineers: ions as regulatorsof morphogenesis during development and regeneration. Tissue Eng.Part A. 14:1461–1468.

36. Chernet, B. T., D. S. Adams, ., M. Levin. 2016. Use of geneticallyencoded, light-gated ion translocators to control tumorigenesis. Onco-target. 7:19575–19588.

37. Spencer Adams, D., J. M. Lemire, ., M. Levin. 2014. Optogeneticsin developmental biology: using light to control ion flux-dependentsignals in Xenopus embryos. Int. J. Dev. Biol. 58:851–861.

38. Adams, D. S., A. S. Tseng, and M. Levin. 2013. Light-activation ofthe Archaerhodopsin Hþ-pump reverses age-dependent loss of verte-brate regeneration: sparking system-level controls in vivo. Biol. Open.2:306–313.

39. Zhao, M., B. Song, ., J. M. Penninger. 2006. Electrical signals con-trol wound healing through phosphatidylinositol-3-OH kinase-g andPTEN. Nature. 442:457–460.

40. Pai, V. P., C. J. Martyniuk,., M. Levin. 2015. Genome-wide analysisreveals conserved transcriptional responses downstream of resting po-tential change in Xenopus embryos, axolotl regeneration, and humanmesenchymal cell differentiation. Regeneration (Oxf). 3:3–25.

41. Tseng, A. S., and M. Levin. 2012. Transducing bioelectric signals intoepigenetic pathways during tadpole tail regeneration. Anat. Rec.(Hoboken). 295:1541–1551.

42. Adams, D. S., S. G. Uzel, ., M. Levin. 2016. Bioelectric signallingvia potassium channels: a mechanism for craniofacial dysmorphogen-esis in KCNJ2-associated Andersen-Tawil Syndrome. J Physiol.594:3245–3270.

43. Masotti, A., P. Uva,., B. Dallapiccola. 2015. Keppen-Lubinsky syn-drome is caused by mutations in the inwardly rectifying Kþ channelencoded by KCNJ6. Am. J. Hum. Genet. 96:295–300.

44. Bando, Y., T. Hirano, and Y. Tagawa. 2014. Dysfunction of KCNK po-tassium channels impairs neuronal migration in the developing mousecerebral cortex. Cereb. Cortex. 24:1017–1029.

45. Pardo, L. A., and W. St€uhmer. 2014. The roles of Kþ channels in can-cer. Nat. Rev. Cancer. 14:39–48.

46. Becchetti, A. 2011. Ion channels and transporters in cancer. 1. Ionchannels and cell proliferation in cancer. Am. J. Physiol. Cell Physiol.301:C255–C265.

47. Djamgoz, M. B., R. C. Coombes, and A. Schwab. 2014. Ion transportand cancer: from initiation to metastasis. Philos. Trans. R. Soc. Lond.B Biol. Sci. 369:20130092.

48. Lobikin, M., B. Chernet,., M. Levin. 2012. Resting potential, onco-gene-induced tumorigenesis, and metastasis: the bioelectric basis ofcancer in vivo. Phys. Biol. 9:065002.

49. Sundelacruz, S., C. Li,., D. L. Kaplan. 2013. Bioelectric modulationof wound healing in a 3D in vitro model of tissue-engineered bone.Biomaterials. 34:6695–6705.

50. Sundelacruz, S., M. Levin, and D. L. Kaplan. 2013. Depolarization al-ters phenotype, maintains plasticity of predifferentiated mesenchymalstem cells. Tissue Eng. Part A. 19:1889–1908.

51. Sundelacruz, S., M. Levin, and D. L. Kaplan. 2008. Membrane poten-tial controls adipogenic and osteogenic differentiation of mesen-chymal stem cells. PLoS One. 3:e3737.

52. Yasuda, T., and D. J. Adams. 2010. Physiological roles of ion channelsin adult neural stem cells and their progeny. J. Neurochem. 114:946–959.

53. Liebau, S., A. Kleger,., S. P. Yu. 2013. Stem cells and ion channels.Stem Cells Int. 2013:238635.

54. Sullivan, K. G., M. Emmons-Bell, and M. Levin. 2016. Physiologicalinputs regulate species-specific anatomy during embryogenesis andregeneration. Commun. Integr. Biol. 9:e1192733.

55. Marsh, G., and H. W. Beams. 1952. Electrical control of morphogen-esis in regenerating Dugesia tigrina. I. Relation of axial polarity tofield strength. J. Cell. Comp. Physiol. 39:191–213.

56. Marsh, G., and H. W. Beams. 1950. Electrical control of growth axisin a regenerating annelid. Anat. Rec. 108:512.

57. Marsh, G., and H. W. Beams. 1949. Electrical control of axial polarityin a regenerating annelid. Anat. Rec. 105:513–514.

58. Marsh, G., and H. W. Beams. 1947. Electrical control of growth po-larity in regenerating Dugesia tigrina. Fed. Proc. 6:163–164.

59. Emmons-Bell, M., F. Durant, ., M. Levin. 2015. Gap junctionalblockade stochastically induces different species-specific head anato-mies in genetically wild-type Girardia dorotocephala flatworms. Int.J. Mol. Sci. 16:27865–27896.

60. Beane, W. S., J. Morokuma,., M. Levin. 2011. A chemical geneticsapproach reveals H,K-ATPase-mediated membrane voltage isrequired for planarian head regeneration. Chem. Biol. 18:77–89.

61. Beane, W. S., J. Morokuma,., M. Levin. 2013. Bioelectric signalingregulates head and organ size during planarian regeneration.Develop-ment. 140:313–322.

Biophysical Journal 112, 2231–2243, May 23, 2017 2241

Durant et al.

62. Garcıa-Quismondo, M., M. Levin, and D. Lobo. 2017. Modelingregenerative processes with membrane computing. Inf. Sci.381:229–249.

63. Durant, F., D. Lobo, ., M. Levin. 2016. Physiological controls oflarge-scale patterning in planarian regeneration: a molecular andcomputational perspective on growth and form. Regeneration (Oxf).3:78–102.

64. De, A., V. S. Chakravarthy, and M. Levin. 2016. A computationalmodel of planarian regeneration. Int. J. Parallel Emergent DistributedSyst. 1–17.

65. Tosenberger, A., N. Bessonov, ., N. Morozova. 2015. A conceptualmodel of morphogenesis and regeneration. Acta Biotheor. 63:283–294.

66. Lobo, D., and M. Levin. 2015. Inferring regulatory networks fromexperimental morphological phenotypes: a computational methodreverse-engineers planarian regeneration. PLOS Comput. Biol.11:e1004295.

67. Levin, M. 2014. Endogenous bioelectrical networks store non-geneticpatterning information during development and regeneration.J. Physiol. 592:2295–2305.

68. Oviedo, N. J., C. L. Nicolas, D. S. Adams, and M. Levin. 2008. Estab-lishing and maintaining a colony of planarians. CSH Protoc. 2008,pdb-prot5053.

69. Nogi, T., and M. Levin. 2005. Characterization of innexin geneexpression and functional roles of gap-junctional communication inplanarian regeneration. Dev. Biol. 287:314–335.

70. Oviedo, N. J., J. Morokuma, ., M. Levin. 2010. Long-range neuraland gap junction protein-mediated cues control polarity duringplanarian regeneration. Dev. Biol. 339:188–199.

71. Pearson, B. J., G. T. Eisenhoffer, ., A. Sanchez Alvarado. 2009.Formaldehyde-based whole-mount in situ hybridization method forplanarians. Dev. Dyn. 238:443–450.

72. Agata, K., Y. Soejima, ., K. Watanabe. 1998. Structure of theplanarian central nervous system (CNS) revealed by neuronal cellmarkers. Zoolog. Sci. 15:433–440.

73. Yazawa, S., Y. Umesono, ., K. Agata. 2009. Planarian hedgehog/patched establishes anterior-posterior polarity by regulating Wntsignaling. Proc. Natl. Acad. Sci. USA. 106:22329–22334.

74. Nakazawa, M., F. Cebria,., T. Gojobori. 2003. Search for the evolu-tionary origin of a brain: planarian brain characterized by microarray.Mol. Biol. Evol. 20:784–791.

75. Cebria, F., C. Kobayashi,., K. Agata. 2002. FGFR-related gene nou-darake restricts brain tissues to the head region of planarians. Nature.419:620–624.

76. Reddien, P. W., A. L. Bermange, ., A. Sanchez Alvarado. 2005.Identification of genes needed for regeneration, stem cell function,and tissue homeostasis by systematic gene perturbation in planaria.Dev. Cell. 8:635–649.

77. Oviedo, N. J., C. L. Nicolas, D. S. Adams, and M. Levin. 2008. Liveimaging of planarian membrane potential using DiBAC43. CSH Pro-toc. 2008, pdb-prot5055.

78. Dexter, J. P., M. B. Tamme, ., E.-M. S. Collins. 2014. On-chipimmobilization of planarians for in vivo imaging. Sci. Rep. 4:6388.

79. Ozkucur, N., K. P. Quinn, ., D. L. Kaplan. 2015. Membrane poten-tial depolarization causes alterations in neuron arrangement and con-nectivity in cocultures. Brain Behav. 5:24–38.

80. Adams, D. S., and M. Levin. 2012. Measuring resting membranepotential using the fluorescent voltage reporters DiBAC43 andCC2-DMPE. Cold Spring Harb. Protoc. 2012:459–464.

81. Adams, D. S., and M. Levin. 2012. General principles for measuringresting membrane potential and ion concentration using fluorescentbioelectricity reporters. Cold Spring Harb. Protoc. 2012:385–397.

82. Epps, D. E., M. L. Wolfe, and V. Groppi. 1994. Characterization of thesteady-state and dynamic fluorescence properties of the potential-sensi-tive dye bis-(1,3-dibutylbarbituric acid)trimethine oxonol (Dibac4(3))in model systems and cells. Chem. Phys. Lipids. 69:137–150.

2242 Biophysical Journal 112, 2231–2243, May 23, 2017

83. Levin, M., T. Thorlin,., M. Mercola. 2002. Asymmetries in Hþ/Kþ-ATPase and cell membrane potentials comprise a very early step inleft-right patterning. Cell. 111:77–89.

84. Vandenberg, L. N., R. D. Morrie, and D. S. Adams. 2011. V-ATPase-dependent ectodermal voltage and pH regionalization are required forcraniofacial morphogenesis. Dev. Dyn. 240:1889–1904.

85. Salo, E., J. F. Abril,., G. Rodrıguez-Esteban. 2009. Planarian regen-eration: achievements and future directions after 20 years of research.Int. J. Dev. Biol. 53:1317–1327.

86. Owlarn, S., and K. Bartscherer. 2016. Go ahead, grow a head! A pla-narian’s guide to anterior regeneration. Regeneration (Oxf). 3:139–155.

87. Gentile, L., F. Cebria, and K. Bartscherer. 2011. The planarian flat-worm: an in vivo model for stem cell biology and nervous systemregeneration. Dis. Model. Mech. 4:12–19.

88. Law, R., and M. Levin. 2015. Bioelectric memory: modeling restingpotential bistability in amphibian embryos and mammalian cells.Theor. Biol. Med. Model. 12:22.

89. Gallaher, J., M. Bier, and J. S. van Heukelom. 2010. First order phasetransition and hysteresis in a cell’s maintenance of the membranepotential—an essential role for the inward potassium rectifiers. Bio-systems. 101:149–155.

90. Pai, V. P., S. Aw, ., M. Levin. 2012. Transmembrane voltage poten-tial controls embryonic eye patterning in Xenopus laevis. Develop-ment. 139:313–323.

91. Mathews, J., and M. Levin. 2016. Gap junctional signaling in patternregulation: physiological network connectivity instructs growth andform. Dev. Neurobiol. 77:643–673.

92. Weingart, R., and F. F. Bukauskas. 1998. Long-chain n-alkanols andarachidonic acid interfere with the Vm-sensitive gating mechanismof gap junction channels. Pflugers Arch. 435:310–319.

93. Juszczak, G. R., and A. H. Swiergiel. 2009. Properties of gap junctionblockers and their behavioural, cognitive and electrophysiological ef-fects: animal and human studies. Prog. Neuropsychopharmacol. Biol.Psychiatry. 33:181–198.

94. Morgan, T. 1898. Experimental studies of the regeneration ofPlanaria maculata. Dev. Genes Evol. 7:364–397.

95. Morgan, T. H. 1904. The control of heteromorphosis in Planaria mac-ulata. Arch. Entwicklungsmech. Org. 17:683–694.

96. Petersen, C. P., and P. W. Reddien. 2008. Smed-betacatenin-1 isrequired for anteroposterior blastema polarity in planarian regenera-tion. Science. 319:327–330.

97. Gurley, K. A., J. C. Rink, and A. Sanchez Alvarado. 2008. b-catenindefines head versus tail identity during planarian regeneration and ho-meostasis. Science. 319:323–327.

98. Sikes, J. M., and P. A. Newmark. 2013. Restoration of anterior regen-eration in a planarian with limited regenerative ability. Nature.500:77–80.

99. Liu, S. Y., C. Selck, ., J. C. Rink. 2013. Reactivating head regrowthin a regeneration-deficient planarian species. Nature. 500:81–84.

100. Hayashi, T., M. Motoishi,., H. Tarui. 2011. A LIM-homeobox geneis required for differentiation of Wnt-expressing cells at the posteriorend of the planarian body. Development. 138:3679–3688.

101. Newmark, P. A., and A. Sanchez Alvarado. 2000. Bromodeoxyuridinespecifically labels the regenerative stem cells of planarians. Dev. Biol.220:142–153.

102. Dahal, G. R., J. Rawson, ., E. Bates. 2012. An inwardly rectifyingKþ channel is required for patterning. Development. 139:3653–3664.

103. Jang, S. S., J. Park, ., S. J. Kim. 2011. Endothelial progenitor cellsfunctionally express inward rectifier potassium channels. Am. J. Phys-iol. Cell Physiol. 301:C150–C161.

104. Klyubin, I. V., K. M. Kirpichnikova,., I. A. Gamaley. 2000. The roleof reactive oxygen species in membrane potential changes in macro-phages and astrocytes. Membr. Cell Biol. 13:557–566.

Bioelectric Control of Target Morphology

105. Levin, M. 2013. Reprogramming cells and tissue patterning viabioelectrical pathways: molecular mechanisms and biomedical oppor-tunities. Wiley Interdiscip. Rev. Syst. Biol. Med. 5:657–676.

106. Levin, M. 2012. Molecular bioelectricity in developmental biology:new tools and recent discoveries: control of cell behavior and patternformation by transmembrane potential gradients. BioEssays. 34:205–217.

107. Forrester, J. V., N. Lois, ., C. McCaig. 2007. The spark of life: therole of electric fields in regulating cell behaviour using the eye as amodel system. Ophthalmic Res. 39:4–16.

108. Adams, D. S., and M. Levin. 2013. Endogenous voltage gradients asmediators of cell-cell communication: strategies for investigatingbioelectrical signals during pattern formation. Cell Tissue Res.352:95–122.

109. Acar, M., J. T. Mettetal, and A. van Oudenaarden. 2008. Stochasticswitching as a survival strategy in fluctuating environments. Nat.Genet. 40:471–475.

110. Chang, H. H., M. Hemberg, ., S. Huang. 2008. Transcriptome-widenoise controls lineage choice in mammalian progenitor cells. Nature.453:544–547.

111. Lobo, D., M. Solano, ., M. Levin. 2014. A linear-encoding modelexplains the variability of the target morphology in regeneration.J. R. Soc. Interface. 11:20130918.

112. Petersen, C. P., and P. W. Reddien. 2009. Wnt signaling and the polar-ity of the primary body axis. Cell. 139:1056–1068.

113. Turing, A. M. 1952. The chemical basis of morphogenesis. Philos.Trans. R. Soc. Lond. B Biol. Sci. 237:37–72.

114. Meinhardt, H. 2007. Computational modelling of epithelialpatterning. Curr. Opin. Genet. Dev. 17:272–280.

115. Grossberg, S. 2013. Adaptive resonance theory: how a brain learns toconsciously attend, learn, and recognize a changing world. NeuralNetw. 37:1–47.

116. Pietak, A., and M. Levin. 2016. Exploring instructive physiologicalsignaling with the bioelectric tissue simulation engine. Front. Bioeng.Biotechnol. 4:55.

117. Pezzulo, G., and M. Levin. 2015. Re-membering the body: applica-tions of computational neuroscience to the top-down control ofregeneration of limbs and other complex organs. Integr. Biol.7:1487–1517.

118. Pezzulo, G., and M. Levin. 2016. Top-down models in biology: expla-nation and control of complex living systems above the molecularlevel. J. R. Soc. Interface. 13:13.

119. Friston, K., M. Levin, ., G. Pezzulo. 2015. Knowing one’s place: afree-energy approach to pattern regulation. J. R. Soc. Interface. 12:12.

Biophysical Journal 112, 2231–2243, May 23, 2017 2243